76JOURNAL OF I.P. EXPERIMENTAL FARIAS ET AL. ZOOLOGY (MOL DEV EVOL) 288:76–92 (2000)

Total Evidence: Molecules, Morphology, and the Phylogenetics of Fishes

1,2 1 3 IZENI P. FARIAS, GUILLERMO ORTÍ, * AND AXEL MEYER 1School of Biological Sciences, University of Nebraska, Lincoln, Nebraska 68588-0118 2Departamento de Biologia, Instituto de Ciências Biológicas, Universidade do Amazonas, 69077-000, Manaus, Amazonas, Brasil 3Department of Biology, University of Konstanz, D-78457 Konstanz, Germany

ABSTRACT We present a most comprehensive phylogenetic analysis of the family Cichlidae. New data analyzed include mitochondrial 16S rRNA sequences and two nuclear loci (Tmo-M27 and Tmo-4C4) for a large taxonomic sampling with emphasis on South American . We also incorporate a published morphological data set for a total evidence analysis. Character congru- ence among mitochondrial (74 taxa) and nuclear data (50 taxa) was high. However, partition- homogeneity tests suggest significant heterogeneity among molecular and morphological data. In agreement with results obtained from molecular data alone, total evidence analysis (1,460 charac- ters for 34 taxa) supports a robust phylogenetic hypothesis for the family Cichlidae that is congru- ent with drift-vicariance events associated with the fragmentation of Gondwana. Our analyses confirm the placement of Malagasy/Indian as the most basal lineages, with a sister-group relationship to the monophyletic African and Neotropical clades. Total evidence suggests that the controversial African Heterochromis is at the base of the African radiation. Among more than 50 Neotropical genera analyzed, Retroculus is identified as the basal taxon, with successive branching of Cichla, Astronotus, geophagines (including crenicichlines) + chaetobranchines, and cichlasomines + heroines. Relative rate tests applied to mitochondrial DNA suggest significantly higher rates of genetic variation in Neotropical than in African taxa, and both mitochondrial and nuclear sequences show that rate heterogeneity among Neotropical lineages is confined to the geophagine cichlids. J. Exp. Zool. (Mol. Dev. Evol.) 288:76–92, 2000. © 2000 Wiley-Liss, Inc.

Cichlids (Cichlidae, Perciformes) have been used trast to African riverine environments, some 550 as model organisms to study a diversity of evolu- cichlid species live in rivers of the New World tionary trends such as parental care, mating sys- (Kullander, ’98) and form an important element tems, sexual selection, and functional morphology of the highly diverse Neotropical ichthyofauna. (see reference Keenleyside, ’91, for reviews on Despite being considerably less speciose than their cichlid biology). Their center of biodiversity is lo- relatives in African lakes, Neotropical cichlids are cated in the East African Great Lakes (Victoria, extremely varied in morphology, behavior, and Malawi, and Tanganyika), that harbor more than ecology (Lowe-McConnell, ’91). Endemism and two-thirds of the estimated 2,000 species in the trophic specialization also are relatively common family. Speciation and adaptive radiations of and, to a certain extent, Neotropical cichlids rep- cichlids evidently have been more pronounced in licate the African lacustrine evolutionary scenario, lacustrine than in riverine environments, prob- albeit at a smaller scale (Stiassny, ’81). Here, we ably facilitated by repeated isolation of subpopu- analyze character congruence among morphologi- lations caused by historical fluctuations of the water level in these lakes (Sturmbauer, ’98; Stiassny and Meyer, ’99). Only 90–100 species live Grant sponsor: University of Nebraska-Lincoln (USA); Grant sponsor: Universidade Federal do Pará (Brazil); Grant sponsor: in African rivers (Greenwood, ’91), about 18 spe- Universidade do Amazonas (Brazil); Grant sponsor: National Sci- cies in Madagascar (Schmidt, ’96), and three spe- ence Foundation; Grant number: DEB-9615178; Grant sponsor: the Deutsche Forschungsgemeinschaft; Grant sponsor: the Fond cies live in southern India and Sri Lanka. Not der Chemischen Industrie; Grant sponsor: University of Konstanz. surprisingly, most phylogenetic studies to date *Correspondence to: Guillermo Ortí, School of Biological Sciences, 314 Manter Hall, University of Nebraska, Lincoln, NE 68588-0118. have focused on the African lacustrine cichlid ra- E-mail: [email protected] diations (e.g., Meyer, ’93; Mayer et al., ’98). In con- Received 4 October 1999; Accepted 27 December 1999 © 2000 WILEY-LISS, INC. CICHLID TOTAL EVIDENCE PHYLOGENY 77 cal data (Kullander, ’98) and three molecular data presumably due to drift-vicariance caused by late sets (one mitochondrial and two nuclear genes) Cretaceous plate-tectonic events. Indian and Mala- used to infer the phylogeny of cichlid fishes. gasy cichlids are considered basal taxa in the fam- The practice of combining independent data sets ily, and form the sister group to the reciprocally into a single phylogenetic analysis continues to monophyletic African and Neotropical clades be debated in the recent literature (Bull et al., (Stiassny, ’91; Zardoya et al., ’96; Streelman et ’93; Eernisse and Kluge, ’93; Chippindale and al., ’98; Farias et al., ’99). Wiens, ’94; Miyamoto and Fitch, ’95; de Queiroz Molecular phylogenetic studies of the whole fam- et al., ’96; Ballard et al., ’98; Kluge, ’98). When ily Cichlidae have been somewhat hampered by data partitions are less heterogeneous than ex- the dearth of Neotropical taxa represented. Sült- pected by sampling error alone, it is generally ac- mann et al. (’95) sequenced anonymous nuclear cepted (e.g., Bull et al., ’93; Chippindale and DNA (DXTU1), Zardoya et al. (’96) microsatellite Wiens, ’94) that all data should be combined to flanking regions (Tmo-M27), Streelman and Karl obtain the best estimate of phylogeny. Tests for (’97) collected data from another single-copy incongruence among data matrices (e.g., Rodrigo nuclear (scn) DNA locus (Tmo-4C4), and Streel- et al., ’93; Farris et al., ’94, ’95) are increasingly man et al. (’98) used both Tmo- scnDNA loci in a used but consensus has yet to emerge regarding combined phylogenetic analysis. However, only how to proceed in cases where significant hetero- two or three species included in these previous geneity is obtained. Many promising alternatives have been suggested recently (e.g., de Queiroz, ’93; studies represented the entire Neotropical cichlid Chippindale and Wiens, ’94; Kluge, ’98; Wiens, radiation of more than 500 species. Although Roe ’98). Additional meaningful outcomes derived from et al. (’97), and Martin and Bermingham (’98) car- examination of incongruence among data parti- ried out extensive taxonomic sampling of Central tions involve an increasing focus on the evolution- American heroines, South American genera re- ary forces underlying character variation among mained poorly represented in their phylogeny of taxa (e.g., Canatella et al., ’98) and assessment of cytochrome b sequences. phylogenetic utility of the characters used. Thus, In a more comprehensive study, Farias et al. procedures used to evaluate patterns of discor- (’98, ’99) sequenced a fragment of the mitochon- dance among data partitions provide not only the drial (mt) 16S rRNA gene for 34 South American basis for deciding how to best conduct phyloge- genera. They identified Neotropical cichlids as a netic analysis (whether to combine or how to com- monophyletic group and suggested that Hetero- bine), but also enlighten our understanding of the chromis and Retroculus are the most basal taxa forces shaping character evolution. of their African and Neotropical clades, respec- Our phylogenetic study is mainly concerned tively. Although the scheme of relationships with the South American cichlid radiation. Previ- among Neotropical genera obtained by this ous analyses of morphological traits have prompted study was highly resolved, the 16S rRNA data the informal subdivision of Neotropical cichlids into (550 bp) provided but a single-locus estimate several suprageneric groups: chaetobranchines, cre- to confidently establish all relationships among nicichlines, geophagines, heroines, and cichlaso- South American and Old World cichlids (Farias mines. Stiassny (’91) suggested three unresolved et al., ’99). In this paper we include all the pre- alternatives for the interrelationship of these groups viously available evidence and present phylo- (her Fig. 1.5). More recently, Kullander (’98) pub- genetic results based on new 16S mtDNA and lished a fully resolved phylogeny and proposed a scnDNA (Tmo-M27 and Tmo-4C4) sequences for new classification for the family Cichlidae, espe- cially for the South American genera, assigning a comprehensive taxonomic sampling of Neotro- new formal names to subfamilies and tribes (for pical cichlids (about 35 additional taxa). We as- more details see: www.nrm.se/ve/pisces/acara). An sess character congruence among molecular unanticipated outcome of this work was the in- data sets and the morphological data published clusion of the enigmatic riverine genus Hetero- by Kullander (’98). Our main goal is to estimate chromis (endemic from the Congo Basin) within relationships among the major groups of cich- an otherwise Neotropical clade (Kullander, ’98). lids, with special emphasis on the South Ameri- Most phylogenetic studies to date have consis- can radiation. We also examine rates and tently identified the major lineages of cichlids con- patterns of evolution in scnDNA and mtDNA forming to their current continental distributions, sequences among the cichlid lineages. 78 I.P. FARIAS ET AL. MATERIALS AND METHODS stepwise additions were performed under maxi- Taxon samples and DNA methods mum parsimony (MP). For all analyses, transi- tions and transversions were weighted equally, The fish taxa included in this study are listed and gaps were treated as “missing” data. Tree in the Appendix. The locality and voucher speci- length (L), consistency indices (CI, excluding un- men information for Neotropical cichlids is avail- informative characters, Kluge and Farris, ’69), and able from Izeni P. Farias. Suprageneric groups (e.g., retention indices (RI, Farris, ’89) are reported in cichlasomines, heroines, etc.) follow common usage each case. Minimum evolution (ME) methods and do not have formal systematic implications (Kidd and Sgaramella-Zonta, ’71) were also ap- (Stiassny, ’91). For a complete list of subfamily and plied to the data, using maximum likelihood dis- tribe names of South American Cichlidae see tances based on the HKY+I+Γ model (Hasegawa Kullander (’98; or www.nrm.se/ve/pisces/acara). For et al., ’85; Yang, ’93; Gu et al., ’95). Parameters all analyses the pomacentrid Abudefduf saxatilis for this model (transition/transversion rate, pro- and the embiotocid Damalichthys vacca were used portion of invariable sites, and gamma shape pa- as outgroups. rameter) were estimated by optimizing the data Total DNA was isolated from ethanol-preserved on the MP trees. tissue by standard proteinase K digestion and phe- The level of confidence in each node of the MP nol/chloroform extraction (Sambrook et al., ’89). and ME trees was assessed using non-parametric PCR amplifications for the nuclear loci Tmo-M27 bootstrapping based on 100 pseudoreplicates (each and Tmo-4C4 followed conditions used by Zardoya with 10 random addition replicates). Bremer sup- et al. (’96) and Streelman and Karl (’97), respec- port (BS) values (Bremer, ’88, ’94) were used as tively. The following internal primers (modified additional measures of node stability under parsi- from Streelman and Karl, ’97) were designed to mony for the total evidence tree (see below). The amplify Tmo-4C4 for heroine cichlids: Tmo-4C4- number of extra steps required to collapse a given ′ ′ F2 5 CGGCCTTCCTAAAACCTCTCATTAAG 3 node was estimated for each clade of interest using ′ and Tmo-4C4-R2 5 GTGCTCCTGGGTGACAAA- constraint-trees constructed in MacClade 3.0 (Mad- ′ GTCTACAG 3 . Conditions used to amplify the dison and Maddison, ’92). The most parsimonious 16S rRNA fragment followed Farias et al. (’99), topologies not satisfying a particular constraint were using primers 16Sar and 16Sbr (Palumbi et al., derived by heuristic searches with replicates (as be- ’91). Double stranded PCR products were se- fore). Tree lengths from unconstrained trees are sub- quenced directly using the BigDye Terminator tracted from tree lengths for constrained trees to cycle sequencing ready reaction kit (Applied determine BS for each node. Biosystems Inc., Foster City, CA) on an automated Previous analyses of 16S rRNA sequences in DNA sequencer (Applied Biosystems 310) follow- cichlid fishes (Farias et al., ’99) suggested hetero- ing manufacturer’s instructions. The nucleotide geneous rates of nucleotide change between Neo- sequence data determined for the present paper tropical and African cichlids. The two-cluster test are deposited in GenBank (accession numbers: of Takezaki et al. (’95) was applied to gauge the 165: AF045842-AF045865, AF048996-AF049019, effect of including additional African and Neotro- AF112577-AF112597, AF112634-AF112642; Tmo- pical taxa in the present study (see Appendix). M27: AF112598-AF112633; Tmo-4C4: AF113060- The test statistic (Z value) represents the devia- AF113095). tion from zero of δ = La – Lb, where La and Lb are the average number of observed substitutions Sequence alignment and per site (branch lengths), from the outgroup to phylogenetic analyses the common ancestor of two monophyletic lineages DNA sequences were aligned using CLUSTALW A and B, respectively. The test is implemented in 1.5 (Thompson et al., ’94). Settings for CLUST- the PHYLTEST program (Kumar, ’96). Constancy ALW were opening gap cost = 20, extending gap of nucleotide substitution rates among Neotropi- cost = 5. Each locus was aligned individually. For cal and African cichlids was tested using the Mala- the Tmo-M27, only the regions flanking the gasy-Indian cichlids as outgroup. The basal microsatellite repeat were used for phylogenetic position of these lineages among cichlids has been analysis, following Zardoya et al. (’96). supported by morphological (Stiassny, ’91) and All phylogenetic analyses were performed us- several molecular data sets (Zardoya et al., ’96; ing PAUP* version 4.0 b2a (Swofford, ’98). Heu- Streelman et al., ’98). In order to minimize the ristic searches with 50 repetitions using random effect of non-independence among distances, a CICHLID TOTAL EVIDENCE PHYLOGENY 79 single representative of each Neotropical genus partitions, rather these tests should be used to was used when more than one species per genus assess the distribution, nature, and extent of con- was available. Relative rate tests are both more flict among data sets.” accurate and more powerful when only the nearest outgroup sequence is used rather than when all RESULTS available outgroup sequences are used (Robinson Patterns of nucleotide divergence et al., ’98). In the present study, the single-nearest outgroup sequence used was Oxylapia for the16S We sequenced the 16S rRNA fragment for 74 data and Etroplus for nuclear loci and the combined taxa. The inclusion of six new African and Neo- molecular data sets. When the relative rate test was tropical taxa in the present study (see Appendix) applied only to the Neotropical lineages, the Afri- did not affect the patterns of nucleotide divergence can genus Heterochromis was used as outgroup. obtained in our previous study (Farias et al., ’99). For each test, new alignments were performed and The final alignment of 16S rRNA sequences re- the model parameters estimated. sulted in 562 bp, of which 276 characters were variable and 206 were phylogenetically informa- Data partitions tive under parsimony. The mean base composi- tion was the same as observed before (Farias et The present study involved the following sets al., ’99), and other estimated parameters were of analyses. The first set included separate very similar to previous results (the transition/ analysis for mitochondrial (16S) and nuclear transversion ratio was 2.42, the proportion of vari- loci (Tmo-M27, Tmo-4C4). The second set in- able sites = 0.34 and gamma shape = 0.50). The cluded a combination of both, mitochondrial and maximum pairwise divergence value observed nuclear sequences (total molecular evidence). In the among all taxa was 0.20 (uncorrected “p” distance) third series of analyses the molecular data (16S + between Teleocichla centrarchus and Paretroplus nuclear loci) were combined with morphological data polyactis. On average, the number of changes per (see Kullander, ’98, for matrix and list of charac- site was 2.8. ters) to reveal the phylogenetic pattern that is sup- Both Tmo-M27 and Tmo-4C4 are suspected to ported by total evidence analysis. Since data for all be protein-coding DNA sequences (Streelman et partitions were not available for all taxa, the taxo- al., ’98). According to Streelman et al., (’98) the nomic composition of each series of analysis is, by Tmo-M27 nucleotide sequence shows significant necessity, somewhat different. matches to RAS-specific guanine nucleotide-re- Following Bull et al. (’93), we first analyzed the leasing factor (RAS-GRF). They also found that different data sets separately and then tested for Tmo-4C4 polypeptide sequence revealed high simi- heterogeneity between data partitions, before com- larity to several regions of TITIN and TITIN-like bining the data in a total evidence analysis. A proteins believed to be related to immunoglobu- simple test, the incongruence length difference lin (IG) domains, which increase the elasticity of (ILD) test, described by Farris et al. (’94, ’95) mea- muscle contraction. In the present work, both sures the significance of incongruence among data Tmo-M27 (302 bp) and Tmo-4C4 (511 bp) se- sets. This test, also known as the partition-homo- quences combined resulted in a total of 813 bp, of geneity test, is implemented in PAUP*4.0 b2a which 400 bp were variable and 250 bp phyloge- (Swofford, ’98). The question of whether “to com- netically informative under parsimony. The mean bine or not to combine” data sets for phylogenetic base composition was 28% A, 26% T, 27% G, and analysis remains controversial (Bull et al., ’93; 19% C. The transition/transversion ratio was 3.19, Eernisse and Kluge, ’93; Chippindale and Wiens, an evident bias in favor of transitions. Other pa- ’94; Miyamoto and Fitch, ’95; Huelsenbeck et al., rameters estimated were proportion of invariable ’96; de Queiroz et al., ’96; Cunningham, ’97; sites = 0.17 and gamma shape = 0.72. From a to- Ballard et al., ’98; Kluge, ’98). We adopt the posi- tal of 50 taxa included in the present work, 36 tion that data sets should be analyzed both sepa- new taxa were sequenced for both nuclear loci. rately and combined simultaneously to potentially The maximum pairwise divergence value observed increase the descriptive efficiency and explanatory among all taxa was 0.21 (uncorrected “p” distance) power of the data (Eernisse and Kluge, ’93). We between Damalichtys vacca and Hemichromis agree with Liu and Miyamoto (’99) that the het- bimaculatus. The average number of changes per erogeneity test “should not be used to justify au- site for the combined nuclear data set was 1.1 (0.9 tomatically the continued separation of character for Tmo-M27 and 1.4 for Tmo-4C4). 80 I.P. FARIAS ET AL. The combined molecular data set comprised a form a monophyletic group strongly supported total of 1,371 bp for 48 taxa (558 bp of 16S rRNA by bootstrap value of 91%; (4) cichlasomines and and 813 bp of Tmo-M27 and Tmo-4C4). Of 653 heroines are sister groups; and (5) Acaronia is variable characters, 432 were phylogenetically in- placed among cichlasomines. formative under parsimony. The estimated tran- sition/transversion ratio was 2.70. The other Total molecular evidence parameters estimated for the HKY+I+Γ model As expected from the overall taxonomic congru- were proportion of invariable sites = 0.29 and ence among nuclear and mtDNA phylogenies gamma shape = 0.54. (Figs. 1 and 2), the partition-homogeneity test in- dicated no significant heterogeneity among the Phylogenetic analyses data partitions (Table 1). The combined molecu- 16S rRNA mitochondrial gene lar evidence for 46 cichlid taxa resulted in 12 equally parsimonious trees (L = 2,032, CI = 0.446, The parsimony analyses of the 16S rRNA data RI = 0.601). A strict consensus of these trees shows resulted in three equally parsimonious trees (L = a trichotomy involving Astronotus, the geopha- 1,566, CI = 0.305, RI = 0.594). Differences among gine+chaetobranchine clade, and the heroine- the three MP trees involve relationships within cichlasomine clade. The ME analysis resulted in the heroine and cichlasomine clades. ME resulted a single tree (score = 1.41, Fig. 3). The topology in a single tree (score = 2.60, Fig. 1). These to- and bootstrap values supporting the MP and ME pologies are consistent with our previous findings trees are highly congruent (Fig. 3), and in agree- (Farias et al., ’99), and support the monophyly of ment with the separate analyses (Figs. 1 and 2). the Cichlidae, the monophyly of the African and The Cichlidae is monophyletic, the Malagasy/In- Neotropical cichlids, and the basal position and dian cichlids are the most basal group in the fam- paraphyly of Malagasy/Indian cichlids. The main ily, and the African and Neotropical cichlids are differences among MP and ME trees involve the monophyletic and sister groups. Relationships placement of the chaetobranchines, and the rela- among African cichlids are consistent in placing tionships among some African taxa. Heterochromis as the most basal African taxon fol- lowed by the other West-African lineages Tylo- Nuclear loci chromis and Hemichromis. In both MP and ME The partition-homogeneity test did not show sig- trees (Fig. 3), African monophyly is supported by nificant heterogeneity (P = 0.86) among the 86% and 90% bootstrap values, respectively. nuclear gene fragments (Table 1), and thus both Among Neotropical cichlids, Retroculus is identi- were combined for further analyses. Phylogenetic fied as the most basal genus together with Cichla analysis of 48 cichlid taxa, based only on the and Astronotus. The reciprocal monophyly of nuclear loci (Tmo-M27 and Tmo-4C4), resulted in cichlasomines, heroines, and geophagines is well 76 equally parsimonious trees (L = 877, CI = supported by the combined data. A close relation- 0.566, RI = 0.711) a strict consensus of which is ship between chaetobranchines (Chaetobranchus+ shown in Figure 2. This tree is similar to the Chaetobranchopsis) and geophagines was found trees obtained by Zardoya et al. (’96) and in all analyses and supported by bootstrap values Streelman et al. (’98) in their analyses of of 79% and 75% in MP and ME trees (Fig. 3). nuclear genes, and shows that African and Neo- tropical cichlids form distinct monophyletic Total evidence clades supported by 67% and 99% bootstrap val- The partition-homogeneity test (Table 1) showed ues, respectively. The combined nuclear loci did significant heterogeneity levels in the comparison not provide sufficient phylogenetic information between molecular and morphological data (P ≤ to confidently resolve all relationships among 0.01). This result was obtained in the face of a South American cichlids. However, important comparatively small number of informative mor- taxonomic congruence with respect to the mt- phological characters with regard to the molecu- DNA tree (Fig. 1 and Farias et al., ’99) includes: lar data. Cunningham (’97) applied different (1) Heterochromis is the most basal genus in- statistical tests to explore the relationship be- cluded in a monophyletic African clade; (2) in tween incongruence and phylogenetic accuracy. He the Neotropical lineage, Retroculus is the most found that with heterogeneity values P ≥ 0.01, basal genus; (3) the geophagines, including combining data might still improve phylogenetic crenicichlines (Crenicichla and Teleocichla), accuracy, and only with P values ≤ 0.01 a com- CICHLID TOTAL EVIDENCE PHYLOGENY 81

Fig. 1. ME tree based on 16S rRNA data obtained using the left and right of the branches are bootstrap values for HKY+I+Γ distances (ME score = 2.60, Tratio = 2.42, Pinvar = MP and ME, respectively. N indicates a bootstrap value be- 0.34, gamma shape = 0.50). Numbers above branches are boot- low 50% for either MP or ME. strap values (only values above 50 are shown). Numbers on 82 I.P. FARIAS ET AL. TABLE 1. P values for partition-homogeneity tests1 monophyly of heroines. Our analysis showed a Comparisons P values close relationship among Heros, Uaru, Mesonauta, Symphysodon, and Pterophyllum, as was sug- Tmo-M27 × Tmo-4C4 loci 0.86 gested by Kullander (’83) and Cichoki (’76). Among 16S rRNA × nuclear loci 0.04 16S rRNA × morphological data 0.01* taxa included in this study, the basal position Nuclear loci × morphological data 0.01* within heroines is occupied by the Hypselecara- Molecular data × morphological data 0.01* Hoplarchus clade. The geophagines are a mono- 1100 random replicates. phyletic group that includes crenicichlines, with *Significant heterogeneity at the 99% confidence level. bootstrap values of 93% and BS values of 9. De- spite low bootstrap values, chaetobranchines are closely related to geophagines in all analyses. bined data matrix may be less accurate than in- In conclusion, for the Neotropical assemblage, dividual data partitions. We follow Cunningham the total evidence results confirm: (1) the place- (’97) and Sullivan (’96) in considering a signifi- ment of Retroculus, Cichla, and Astronotus as cance threshold of 0.05 to be too conservative for basal lineages. In contrast to Stiassny (’91) and the homogeneity test. Furthermore, combining Kullander (’98) Cichla is not closely related to data sets may still be advantageous despite high crenicichlines; (2) the monophyly of sister groups levels of character incongruency, since the overall heroines and cichlasomines, with Acaronia in- accuracy of the combined data may be increased cluded in the cichlasomine group; (3) the mono- by a larger number of characters applied to parts phyly of geophagines including the crenicichline of the tree unaffected by the mismatch (Wiens, group; and (4) the close relationship among ’98). Furthermore, by exploring taxon sampling chaetobranchines and geophagines. in relation to heterogeneity tests, we attempt to identify the portions of the tree most affected by Rates of molecular evolution character incongruence (see Discussion). The 16S rRNA data resulted in evident differ- All molecular and morphological character data ences in branch lengths in the ME tree (Fig. 1). available were subsequently combined for a total As reported by Farias et al. (’99), higher rates of evidence analysis (16S + nuclear loci + morpho- evolution in Neotropical lineages, and in particu- logical data; 1,460 characters for 34 taxa). Figure lar among geophagine cichlids, were detected 4 shows a strict consensus tree of three equally based on the distribution of the genetic distances parsimonious trees obtained (L = 2,155, CI = (Table 2). The relative rate test Z-statistic (Z = 0.368, RI = 0.485). Differences among the MP trees 4.1342) rejected rate constancy (P < 0.05) among only involved the placement of Heros, Uaru, and Neotropical and African lineages, showing that the Mesonauta within heroines. The combined data average distance among lineages was significantly set provided increased resolution and higher boot- different in both continental faunas. Even when strap support compared to any of the individual the geophagine group was excluded from the Neo- data partitions. In agreement with our previous tropical clade, the average rate of substitution results, total evidence clearly shows that African among Neotropical and African cichlids was sig- and Neotropical cichlids are distinct monophyletic nificantly different (Z = 2.5661). Only when lacus- clades supported by bootstrap values of 92% and trine cichlids (short branch lengths) in Africa and 100% and BS values of 8 and 15, respectively. the geophagines (long branch lengths) were ex- Etroplus is placed as the most basal cichlid and cluded the relative rate test was non-significant Heterochromis is placed as the basal lineage of (Table 2). The relative rate test applied only to the African clade followed by Tylochromis. the Neotropical clade showed significantly differ- Among the Neotropical cichlids, Retroculus is ent rates among geophagines compared with the the most basal genus followed by Cichla and rest of the Neotropical taxa (Z = 2.7996, P < 0.05). Astronotus. Two major lineages are observed, one A separate analysis using 16S rRNA data per- formed by the cichlasomine and heroine groups formed only on Old World cichlids confirmed a dif- and a second by the chaetobranchine and geo- ference in the branch lengths of some lineages phagine groups. The monophyly of cichlasomines among African taxa in the ME tree (Fig. 1). An is supported by a bootstrap value of 84% and BS unrooted tree (obtained by ME, result not shown) value of 4. The relationships among genera were suggested that long branches are found mainly the same as observed in all analyses. Bootstrap in the basal riverine African cichlids and also in values of 89% and BS values of 4 supported the some taxa from Lake Tanganyika. A relative rate CICHLID TOTAL EVIDENCE PHYLOGENY 83

Fig. 2. Strict consensus of 76 MP trees from nuclear loci Numbers above branches are bootstrap values (only values data (Tmo-M27 + Tmo-4C4) (L = 877, CI = 0.566, RI = 0.711). above 50% are shown). 84 I.P. FARIAS ET AL.

Fig. 3. ME tree from total molecular evidence (16S rRNA right of the branches are bootstrap values for MP and ME + nuclear loci) obtained using HKY+I+Γ distances (score = (only values above 50% are shown). N indicates values below 1.41, Tratio = 2.70, Pinvar = 0.29). Numbers on the left and 50% for either MP or ME tree. CICHLID TOTAL EVIDENCE PHYLOGENY 85 test among riverine and lacustrine African cichlids additional partition-homogeneity tests of mo- (with Oxilapia as outgroup) was performed (Table lecular and morphological subsets, using the 34- 2). The analysis revealed a surprising difference taxon total evidence data set. Removal of these in branch lengths between riverine (La = 0.0618) taxa individually did not result in change of P and lacustrine (Lb = 0.0315) cichlids, rejecting rate value (for all tests P = 0.01). Higher P values constancy (P < 0.05) with a significant average were observed only when we removed two or distance value (Z = 2.8521). more taxa, but never was P > 0.06 for any combi- Despite observing considerable difference in nation. The largest change was obtained by the branch lengths in the ME tree based on the Tmo- simultaneous removal of Heterochromis, Acaronia, M27 data and smaller differences in the Tmo-4C4 and Crenicichla and by these three plus Cichla tree (results not shown), significant heterogene- (for both cases P = 0.052, with 500 replications). ity was not detected between Neotropical and Af- Given that these four taxa alone account for the rican cichlids for nuclear loci (Table 2). However, major taxonomic discordance among morphologi- when the test was applied only to the Neotropical cal and molecular trees while all other aspects of assemblage, Tmo-4C4 data showed significant dif- the phylogeny are mostly congruent, it seems jus- ferences (P < 0.05) among geophagines and the tified to set higher stringency standards (i.e., P < rest of the Neotropical taxa, confirming rate het- 0.01 or smaller) to determine significant levels of erogeneity in both mitochondrial and nuclear DNA character incongruency using partition-homogene- among the Neotropical lineages. ity tests, as suggested by Cunningham (’97) and Sullivan (’96). DISCUSSION Phylogenetic placement of the African genus Taxonomic and character congruence Heterochromis (subfamily Heterochromidinae) In a vast majority of studies, data from single among the basal lineages of the Neotropical clade mitochondrial genes are used to address system- (Fig. 5) is not supported by molecular data. Based atic questions. However, single-gene reconstruc- on 16S rRNA sequences alone, Farias et al. (’99) tions may lead to inconclusive discrimination of suggested that Heterochromis constitutes a basal species trees among gene trees (Moore, ’95). Only branch of the African clade, an observation sup- if one or more nuclear genes independently re- ported by additional taxa and multilocus infor- solve the same topology as mtDNA, inference of mation presented here. Kullander (’98) placed the species tree rather than gene trees is strongly Heterochromis among Neotropical cichlids (see corroborated. Congruence among different data Fig. 5) based on the following three morpho- partitions has arguably provided the strongest evi- logical characters: (i) short anterior arm of dence that a particular phylogenetic estimate is epibranchial 1 (his character 2); (ii) interdigi- accurate (Penny and Hendy, ’86; Swofford, ’91). tating suture between the vomerine shaft and the Comparison of Figures 1–4 shows that the con- parasphenoid bar (his character 37); and (iii) an- cordant topologies obtained for the major lineages terior palatoethmoid ligament present (his char- of cichlids in the present work are likely to be a acter 57). Our results, supported by total evidence, reflection of organismal phylogeny rather than suggest that these morphological characters simply gene genealogies. Our phylogenetic analy- shared among Heterochromis and Neotropical ses of mitochondrial (16S rRNA) and nuclear loci cichlids may have originated independently. (Tmo-M27, Tmo-4C4) combined with morphology Homoplasious phenotypes attest to the strength (from Kullander, ’98) constitutes the most com- of selection or, alternatively, internal constraints plete phylogenetic analysis of South American to sculpt similar phenotypes in response to simi- lar selection pressures and to re-express ancient, cichlids to date. retained, developmental programs (Meyer, ’99). Our major systematic findings are compared to Quicke and Belshaw (’99) demonstrated eloquently Kullander’s (’98) hypothesis in Figure 5. Major dis- how convergence among certain morphological cordance between the morphological and the to- characters resulting from a shared life history tal evidence results involve the relationships of strategy may result in misleading phylogenetic Heterochromis, Cichla, Crenicichla, Chaeto- reconstruction of the evolution of endoparasitism branchus, and Acaronia. To gauge the effect of among braconid wasps. Stiassny (’92) used the taxon sampling on the assessment of character family Cichlidae as one example to suggest that incongruence, these taxa were removed indi- much character incongruence present in morpho- vidually and in all possible combinations, for logical evolution is the result of taxic atavism (a 86 I.P. FARIAS ET AL.

Figure 4. CICHLID TOTAL EVIDENCE PHYLOGENY 87

Fig. 5. Comparison among phylogenetic hypotheses from Kullander (’98) and the total evidence approach proposed in the present work.

back mutation from a derived character state to an ancestral state). It remains to be seen if these Fig. 4. Strict consensus of three MP trees from Total Evi- three morphological characters have been influ- dence (16S + nuclear loci + morphological data, 1,460 char- enced by such processes. Although it is well-known acters) (L = 2,155, CI = 0.368, RI = 0.485). The numbers above branches are bootstrap values (only values above 50 that molecular data are not free of homoplasy (CI are shown). The numbers below internodes correspond to values for our molecular data range from 0.37 to branch support values. 0.51), the fact that three independent loci are phy- 88 I.P. FARIAS ET AL.

TABLE 2. Relative rate tests

Cluster Sequence A × BLaLbδ Z

16S rRNA La = Neotropical × Lb = African 0.109 0.046 0.062 4.13* La = Neotropical (excluding geophagines) × Lb = African 0.088 0.051 0.036 2.56* La = Geophagines × Lb = Rest Neotropical 0.103 0.065 0.037 2.80* La = River Afr × Lb = Lake Afr cichlids 0.062 0.031 0.030 2.85* La = River Afr × Lb = Neotropical cichlids 0.111 0.073 0.037 2.18* La = River Afr × Lb = Neotropical (excluding geophagine) 0.083 0.070 0.013 0.85 Tmo-M27 La = Neotropical × Lb = African 0.061 0.035 0.025 1.50 La = Geophagines × Lb = Rest Neotropical 0.030 0.015 0.015 1.89† Tmo-4C4 La = Neotropical × Lb = African 0.060 0.090 0.030 1.69 La = Geophagines × Lb = Rest Neotropical 0.050 0.020 0.031 4.20* Nuclear loci La = Neotropical × Lb = African 0.057 0.064 0.007 0.59 La = Geophagines × Lb = Rest Neotropical 0.039 0.016 0.023 4.64* *Rate constancy is rejected at the 95% level. †Value close to 1.96 required to reject the rate constancy at the 95% level (Kumar, ’96).

logenetically concordant argues against significant Neotropical geophagine lineage were excluded systematic bias in the DNA data (Swofford et al., from the analysis. Tests applied to both mitochon- ’96) that could account for nonhistorical hierar- drial and nuclear sequences confirmed that rate chical signal (Naylor and Brown, ’98). heterogeneity within the Neotropical clade is con- Combining data is currently controversial and fined to the geophagine group. When applied to total evidence (character congruence) is only one African cichlids alone (using Oxylapia as out- of several methods that have been suggested (re- group), only 16S rRNA sequences showed signifi- viewed by Miyamoto and Fitch, ’95). The total evi- cant rate heterogeneity among riverine and dence analyses may be preferred not because it is lacustrine African cichlids. more likely to find the unknowable “true” tree, The amount of genetic variation within lineages however, rather because it searches, in a single may be an indication of their age, assuming more hypothesis, to account for all of the relevant data or less constant mutation rates and demography. within a parsimony framework (e.g., Kluge, ’89). High levels of genetic variation have been ob- served for mitochondrial genes in cichlids from Rate comparison Lake Tanganyika when compared to species flocks The basic idea of the relative-rate test is to com- from Lake Malawi and Victoria (Meyer et al., ’90; pare distances between two lineages since they Sturmbauer and Meyer, ’92). A similar observa- last shared a common ancestor. Increasing the tion was reported by Takahashi et al. (’98) using number of ingroup sequences improves both ac- SINEs data, describing a relatively long basal curacy and power of relative rate tests from a sta- branch for the common ancestor of the extant spe- tistical point of view. According to Robinson et al. cies of the Lamprologini, an endemic group of (’98) this is obtained by using as many distantly cichlids from Lake Tanganyika. Several studies related ingroup sequences as possible, and by us- have shown that the whole cichlid flock in Lake ing the single nearest outgroup sequence alone. Tanganyika originated from a few ancient and ge- Thus, using a closely related outgroup makes the netically divergent lineages derived from riverine statistical test of rate constancy more powerful. ancestors (Meyer, ’93; Sturmbauer et al., ’94; Previously, we found rate heterogeneity among Nishida, ’97; Takahashi et al., ’98). Although rep- Neotropical and African cichlids using 16S rRNA resentation of African taxa diversity is somewhat sequences (Farias et al., ’99). In the present study, limited in our study, our results also support pre- by increasing the representation of African taxa, vious suggestions that the West African riverine those results were confirmed. However, different cichlids are the basal lineages among African rates between Neotropical and African taxa were cichlids. Although highly speciose in lakes, cichlids significant only for the 16S rRNA mitochondrial in rivers seem likely to preserve a larger fraction gene. The relative rate test failed to find signifi- of the total genetic variation represented in the cant rate differences (Table 2) only when lacus- family. Kullander (’98) suggested that riverine trine cichlids in the African lineage and the cichlid faunas are more likely to conserve ances- CICHLID TOTAL EVIDENCE PHYLOGENY 89 tral patterns of character distribution. Compari- Bremer K. 1994. Branch support and tree stability. Cladis- son of riverine cichlids from Africa and the tics 10:295–304. Bull JJ, Huelsenbeck P, Cunningham CW, Swofford DL, Neotropics with their well-studied lacustrine rela- Waddell PJ. 1993. Partitioning and combining data in phy- tives may help understand this pattern. Lamen- logenetic analysis. Syst Biol 42:384–397. tably, their ecology, genetics, and evolution have Cannatella DC, Hillis DM, Chippindale PT, Weigt L, Rand been poorly characterized in contrast to that of AS, Ryan MJ. 1998. Phylogeny of frogs of the Physalaemus cichlids from the Great African lakes (Greenwood, pustulosus species group, with an examination of data in- congruence. Syst Biol 47:311–335. ’91; Lowe-McConnell, ’91; Mayer et al., ’98). Chippindale PT, Wiens JJ. 1994. Weighting, partitioning, and The present results suggest that the rate of mo- combining characters in phylogenetic analysis. Syst Biol lecular evolution among Neotropical lineages may 43:278–287. not be uniform. Rate heterogeneity can vary Cichoki F. 1976. Cladistic history of cichlid fishes and repro- among phyletic lineages, genes, and gene regions ductive strategies of the American genera Acarichthys, Biotodoma, and Geophagus. Ph.D. thesis, University of within the same genome (Lopez et al., ’97). Fur- Michigan, Ann Arbor. thermore, rate heterogeneity among two lineages Cunningham CW. 1997. Can tree incongruence tests pre- does not guarantee a molecular clock within each dict when data should be combined? Mol Biol Evol lineage (Robinson et al., ’98). Differences in rates 14:733–740. of molecular clocks might be related to body size, de Queiroz A. 1993. For consensus (sometimes). Syst Biol 42:368–372. generation time, metabolic rate, DNA repair de Queiroz A, Donoghue MJ, Kim J. 1996. Separate versus mechanisms and effective population size, and of combined analysis of phylogenetic evidence. Annu Rev Ecol course constraints on function of the gene prod- Syst 26:657–681. uct (Wu and Li, ’85; Martin and Palumbi, ’93; Eernisse DJ, Kluge AG. 1993. Taxonomic congruence versus Moore, ’95). Future work should test which of total evidence, and amniote phylogeny inferred from fos- sils, molecules, and morphology. Mol Biol Evol 10:1170–1195. these hypotheses best explains the rate heteroge- Farris JS. 1989. The retention index and rescaled consistency neity observed among Neotropical and African index. Cladistics 5:417–419. cichlids. We hope that our demonstration of rate Farris JS, Källersjö M, Kluge AG, Bult C. 1994. Testing sig- heterogeneity among Neotropical cichlids may nificance of incongruence. Cladistics 10:315–319. stimulate future studies correlating life history Farris JS, Källersjö M, Kluge AG, Bult C. 1995. Constructing a significance test for incongruence. Syst Biol 44:570–572. traits with rates and patterns of molecular evolu- Farias IP, Schneider H, Sampaio I. 1998. Molecular phylog- tion in cichlid fishes. eny of Neotropical cichlids: the relationships of cichlaso- mines and heroines. In: Malabarba LR, Reis RE, Vari RP, ACKNOWLEDGMENTS Lucena ZM, Lucena CAS, editors. Phylogeny and classifi- We thank the many colleagues who provided tis- cation of Neotropical fishes. Porto Alegre, Brasil: Edipucrs. sue samples and species identification: A. Basolo, p 499–508. Farias IP, Ortí G, Sampaio I, Schneider H, Meyer A. 1999. C. Silva, E. Feldberg, E. Ferreira, I. Koslowski, J. Mitochondrial DNA phylogeny of the family Cichlidae: mono- Alves-Gomes, J. Porto, J. Zuanon, M. Garcia, R. phyly and fast molecular evolution of the Neotropical as- Reis, S. Kullander, T. Ploesch and W. Staeck. We semblage. J Mol Evol 48:703–711. thank T. Hrbek, W. Salzburger and members of Greenwood PH. 1991. Speciation. In: Keenleyside MHA, edi- the Ortí lab for providing useful comments on ear- tor. Cichlid fishes—behavior, ecology, and evolution. Lon- don: Chapman and Hall. p 86–102. lier versions of the manuscript. This work was Gu X, Fu Y-X, Li W-H. 1995. Maximum likelihood estimation supported in part by University of Nebraska-Lin- of the heterogeneity of substitution rate among nucleotide coln (USA), Universidade Federal do Pará (Bra- sites. Mol Biol Evol 12:546–557. zil), and Universidade do Amazonas (Brazil). IPF Hasegawa M, Kishino H, Yano T. 1985. Dating of the hu- is a Ph.D. fellow from CAPES (Brazil). We also man-ape splitting by a molecular clock of mitochondrial DNA. J Mol Evol 22:160–174. acknowledge support from the National Science Huelsenbeck JP, Bull JJ, Cunningham CW. 1996. Combining Foundation (DEB-9615178), the Deutsche Fors- data in phylogenetic analysis. TREE 11:152–158. chungsgemeinschaft, the Fond der Chemischen Keenleyside MHA, editor. 1991. Cichlid Fishes—behavior, ecol- Industrie, and University of Konstanz to A.M. ogy, and evolution. London: Chapman and Hall. Kidd KK, Sgaramella-Zonta LA. 1971. Phylogenetic analy- LITERATURE CITED sis: concepts and methods. Am J Hum Genet 23:235–252. Kluge AG. 1989. A concern for evidence and a phylogenetic Ballard JWO, Thayer MK, Newton AF, Grismer ER. 1998. Data hypothesis of relationships among Epicrates (Boidae, sets, partitions, and characters: philosophies and procedures Serpentes). Syst Zool 38:7–25. for analyzing multiple data sets. Syst Biol 47:367–396. Kluge AG. 1998. Total evidence or taxonomic congruence: Cla- Bremer K. 1988. The limits of amino acid sequence data distics or consensus classification. Cladistics 14:151–158. in angiosperm phylogenetic reconstruction. Evolution Kluge AG, Farris JS. 1969. Quantitative phyletics and the 42:795–803. evolution of anurans. Syst Zool 18:1–32. 90 I.P. FARIAS ET AL.

Kullander SO. 1983. A revision of the South American cichlid sitivity of the relative -rate test to taxonomic sampling. Mol genus Cichlasoma. Stockholm: Swedish Museum of Natu- Biol Evol 15:1091–1098. ral History. Rodrigo AG, Kelly-Borges M, Bergquist PR, Bergquist PL. Kullander SO 1998. A phylogeny and classification of the 1993. A randomization test of the null hypothesis that two South American Cichlidae (Teleostei: Perciformes). In: cladograms are sample estimates of a parametric phyloge- Malabarba LR, Reis RE, Vari RP, Lucena ZM, Lucena CAS, netic tree. N Z J Bot 31:257–268. editors. Phylogeny and classification of Neotropical fishes. Roe KL, Conkel D, Lydeard C. 1997. Molecular systematic of Porto Alegre, Brasil: Edipucrs. p 461–498. Middle American cichlid fishes and evolution of trophic-types Kumar S. 1996. PHYLTEST: phylogeny hypothesis testing soft- in ‘Cichlasoma (Amphilophus)’ and ‘C. (Thorichthys)’. Mol ware. University Park: Pennsylvania State University. Liu FR, Miyamoto MM. 1999. Phylogenetic assessment of Phylogenet Evol 7:366–376. molecular and morphological data for eutherian mammals. Sambrook J, Fritsch EF, Maniatis T. 1989. Molecular clon- Syst Biol 48:54–64. ing: a laboratory manual, 2nd ed. New York: Cold Spring Lopez JV, Culver M, Stephens C, Johnson WE, O’Brien SJ. Harbor Laboratory Press. 1997. Rates of nuclear and cytoplasmic mitochondrial DNA Schmidt W and ag editorial team. 1996. La grandle Ile. Aqua sequence divergence in mammals. Mol Biol Evol 14:277–286. Geographia 13:5–36. Lowe-McConnell RH. 1991. Ecology of cichlids in South Ameri- Stiassny MLJ. 1981. The phyletic status of the family can and African waters, excluding the African Great Lakes. Cichlidae (Pisces: Perciformes): a comparative anatomical In: Keenleyside MHA, editor. Cichlid fishes—behavior, ecol- investigation. Neth J Zool 31:275–314. ogy, and evolution. London: Chapman and Hall. p 60–85. Stiassny MLJ. 1991. Phylogenetic intrarelationships of the Maddison WP, Maddison DR. 1992. MacClade: analysis of phy- family Cichlidae. In: Keenleyside MHA, editor. Cichlid logeny and character evolution, version 3.0. Sunderland, fishes—behavior, ecology, and evolution. London: Chapman MA: Sinauer. and Hall. p 1–31. Martin AP, Bermingham E. 1998. Systematics and evolution Stiassny MLJ. 1992. Atavism, phylogenetic character rever- of lower Central American cichlids inferred from analysis of sals, and the origin of evolutionary novelties. Neth J Zool cytochrome b gene sequence. Mol Phylogenet Evol 9:192–203. 42:260–276. Martin AP, Palumbi SR. 1993. Body size, metabolic rate, Stiassny MLJ, Meyer A. 1999. Cichlids of rift lakes. Sci Am generation time, and the molecular clock. Proc Natl Acad 280:64–69. Sci USA 90:4087–4091. Streelman JT, Karl SA. 1997. Reconstructing labroid evolu- Mayer WE, Tichy H, Klein J. 1998. Phylogeny of African tion with single-copy nuclear DNA. Proc R Soc Lond B cichlid fishes as revealed by molecular markers. Heredity 264:1011–1020. 80:702–714. Streelman JT, Zardoya R, Meyer A, Karl SA. 1998. Multi- Meyer A. 1993. Phylogenetic relationships and evolutionary locus phylogeny of cichlid fishes (Pisces: Perciformes): evo- processes in East African cichlid fishes. Trend Ecol Evol lutionary comparison of microsatellite and single-copy 8:279–284. nuclear loci. Mol Biol Evol 15:798–808. Meyer A. 1999. Homology and homoplasy: the retention of ge- Sturmbauer C. 1998. Explosive speciation in cichlid fishes of netic programmes. In: Novatis Foundation Symposium 222. the African Great Lakes: a dynamic model of adaptive ra- Homology. England: John Wiley & Sons Ltd. p 141–157. diation. J Fish Biol 53:18–36. Meyer A, Kocher TD, Basasibwaki P, Wilson AC. 1990. Mono- Sturmbauer C, Meyer A. 1992. Genetic divergence, specia- phyletic origin of Lake Victoria cichlid fishes suggested by tion and morphological stasis in a lineage of African cichlid mitochondrial DNA sequences. Nature Lond 347:550–553. fishes. Nature Lond 359:578–581. Miyamoto MM, Fitch WM. 1995. Testing species phylogenies Sturmbauer C, Verheyen E, Meyer A. 1994. Mitochondrial and phylogenetic methods with congruence. Syst Biol phylogeny of the Lamprologini, the major substrate spawn- 44:64–76. ing lineage of cichlid fishes from Lake Tanganyika in East- Moore WS. 1995. Inferring phylogenies from mtDNA varia- ern Africa. Mol Biol Evol 11:691–703. tion: mitochondrial-gene trees versus nuclear-gene trees. Sullivan J 1996. Combining data with different distributions of among-site variations. Syst Biol 45:375–380. Evolution 49:718–726. Sültmann H, Mayer WE, Figueroa F, Tichy H, Klein J. 1995. Naylor GJP, Brown WM. 1998. Amphioxus mitochondrial Phylogenetic analysis of cichlid fishes using nuclear DNA DNA, phylogeny, and the limits of inference based markers. Mol Biol Evol 12:1033–1047. on comparisons of sequences. Syst Biol 47: 61–76. Swofford DL. 1991. When are phylogeny estimates from mo- Nishida M. 1997. Phylogenetic relationships and evolution of lecular and morphological data incongruent? In: Miyamoto tanganyikan cichlids: a molecular perspective. In: Kawanabe MM, Cracraft J, editors. Phylogenetic analysis of DNA se- H, Hori M, Nagoshi M, editors. Fish communities in Lake quences. New York: Oxford University Press. p 295–333. Tanganyika. Kyoto, Japan: Kyoto University Press. p 1–23. Swofford DL. 1998. PAUP* Phylogenetic analysis using par- Palumbi S, Martin A, Romano S, McMillian WO, Stice L, simony and other methods, Version 4.0b. Sunderland, MA: Grabowski G. 1991. The simple fool’s guide to PCR. Hono- Sinauer. lulu: University of Hawaii. Swofford DL, Olsen GJ, Waddell PJ, Hillis DM. 1996. Phylo- Penny D, Hendy MD. 1986. Estimating the reliability of evo- genetic inference. In: Hillis DM, Moritz C, Mable BK, edi- lutionary trees. Mol Biol Evol 3:403–417. tors. Molecular Systematics. Second edition. Sunderland, Quicke DLJ, Belshaw R. 1999. Incongruence between morpho- MA: Sinauer. p 407–514. logical data sets: an example from the evolution of endopara- Takahashi K, Terai Y, Nishida M, Okada N. 1998. A novel sitism among parasitic wasps (Hymenoptera: Braconidae). family of Short Interspersed Repetitive Elements (SINEs) Syst Biol 48:436–454. from cichlids: the patterns of insertion of SINEs at ortholo- Robinson M, Gouy M, Gautier C, Mouchiroud D. 1998. Sen- gous loci support the proposed monophyly of four major CICHLID TOTAL EVIDENCE PHYLOGENY 91

groups of cichlid fishes in Lake Tanganyika. Mol Biol Evol Wu CI, Li WH. 1985. Evidence for higher rates of nucleotide 15:391–407. substitution in rodents than in man. Proc Natl Acad Sci Takezaki N, Rzhetsky A, Nei M. 1995. Phylogenetic test of USA 82:1741–1745. the molecular clock and linearized trees. Mol Biol Evol Yang Z. 1993. Maximum likelihood estimation of phylogeny 12:823–833. from DNA sequences when substitution rates differ over Thompson JD, Higgins DG, Gibson TJ. 1994. CLUSTAL W: sites. Mol Biol Evol 10:1396–1401. improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position specific gap Zardoya R, Vollmer DM, Craddock C, Streelman JT, Karl SA, penalties, and weight matrix choice. Nucleic Acids Res Meyer A. 1996. Evolutionary conservation of microsatellite 22:4673–4680. flanking regions and their use in resolving the phylogeny Wiens JJ. 1998. Combining data sets with different phyloge- of cichlid fishes (Pisces: Perciformes). Proc R Soc Lond B netic histories. Syst Biol 47:568–581. 263:1589–1598.

APPENDIX. Taxa of fishes included in this study1

Taxa 16S rRNA Tmo-M27 Tmo-4C4

Embiotocidae Cymatogaster aggregata 3—— Damalichthys vacca 412 Pomacentridae Abudefduf saxatilis 312 Cichlidae Etroplinae (Malagasy and India) Etroplus maculatus 312 Oxylapia polleni 3—— Paretroplus polyactis 312 Ptychochromoides betsileanus 3—— Ptychochromis oligocantus 3—— Pseudocrenilabrinae (Africa) Astatoreochromis alluaudi 412 Astatotilapia callipitera 412 Boulengerochromis microlepis 312 Chalinochromis brichardi 3—— Chromidotilapia sp. 3 — — Haplochromis sp. 3 — — Hemichromis bimaculatus 312 Labidochromis caeruleus 412 Melanochromis sp. 4 — — Oreochromis niloticus 3—— Pelvicachromis pulcher —12 Serranochromis robustus 312 Thysochromis sp. 3 — — Tilapia sp. 3 — — Tropheus moorii 412 Tylochromis polylepis 412 Heterochromidinae (Africa) Heterochromis multidens 344 Retroculinae (South Africa) Retroculus sp. 3 4 4 Retroculus xinguensis 344 Cichlinae (South America) Cichlini Cichla sp. 4 4 4 Cichla monoculus 3—— Cichla orinocensis 344 Cichla temensis 3—— Crenicichlini Crenicichla sp. 3 — — Crenicichla lugubris 3—— Crencichla regani 344 Teleocichla centrarchus 344 Teleocichla cinderella 3—— 92 I.P. FARIAS ET AL.

APPENDIX. (continued)

Taxa 16S rRNA Tmo-M27 Tmo-4C4

Astronotinae (South America) Astronotini Astronotus crassipinnis 3—— Astronotus ocellatus 312 Chaetobranchopsis orbicularis —44 Chaetobranchus flavescens 3—— Chaetobranchus semifasciatus 344 Geophaginae (South America) Acarichthyini Acarichthys heckelii 344 Guianacara sp. 3 4 4 Crenicaritini Biotoecus sp. 4 — — Crenicara sp. — 4 4 sp. 1 3 — — Apistogramma sp. 2 3 4 4 Biotodoma wavrini 344 Geophagus sp. 4 4 4 Geophagus altifrons 344 Geophagus argyrostictus 444 Geophagus proximus 3—— ‘Geophagus’ brasiliensis 344 Gymnogeophagus gymnogenys 344 Gymnogeophagus labiatus 3—— Mikrogeophagus altispinosus 344 Satanoperca acuticeps 3—— Satanoperca jurupari 3—— Taeniacara candidi 344 Cichlasomatinae (South America) Acaroniini (South America) Acaronia sp. 3 — — Acaronia nassa 344 Cichlasomatini (South America) Aequidens michaeli 344 ‘Aequidens’ sp. 3 — — Bujurquina sp. 3 4 4 Cichlasoma amazonarum 344 Laetacara curviceps 344 Nannacara sp. 3 — — Heroini (South America) Caquetaia spectabilis 344 Heros sp. 3 4 4 Hoplarchus psittacus 344 Hypselecara coryphaenoides 344 Mesonauta insignis 344 Pterophyllum scalare 344 Symphysodon aequifasciatus 344 Uaru amphiacanthoides 344 Heroini (Central America) ‘Cichlasoma’ octofasciatum 444 Petenia splendida 444 Amphilophus citrinellu 3—— Archocentrus nigrofasciatus 3—— 1Taxonomy follows Kullander (’98). Genera between quotation marks are not valid for the species indicated (e.g., ‘Aequidens,’ ‘Geophagus,’ ‘Cichlasoma’) but are used for lack of a better name. Voucher specimen and locality information is available from I.P.F. Numbers identify the source of the sequence data analyzed in the present paper. 1: Tmo-M27 data from Zardoya et al. (’96); 2: Tmo-4C4 data from Streelman and Karl (’97); 3: 16S rRNA data from Farias et al. (’99); 4: new data determined in present study.